When Students Wrote Better English Than They Could Explain Lesson from Using AI in the EFL Classroom

By Fransiska Karolina Wilbiasri Ine, S.Pd, Mahasiswa Program Studi Magister Pendidikan Bahasa Inggris Universitas Negeri Semarang

A student submitted a paper with remarkable fluency. The sentence structure was correct, the vocabulary was diverse, and the layout was organized. For a student who usually had difficulty constructing sentences, this progress seemed almost extraordinary. When I first experienced this while teaching through the Kampus Mengajar program, I was truly surprised. For a moment, I felt that this was what progress was all about. Then, I asked a simple question in class. “Can you explain the reason behind the use of this tense in the paragraph you wrote?” But the student was silent. It was then that I realized that artificial intelligence technology was already present in my English class before I had really decided how to deal with it. Even though I was teaching in a rural area at the time, and AI was not yet very popular, but I have to admit that the students were quite proficient in using AI despite the limited network coverage in their area.

Curiosity Meets Reality

Like most English teachers today, I did not start using AI tools with a well-defined teaching plan. What triggered me was curiosity mixed with necessity. Students have begun discussing ChatGPT, Grammarly, and various other writing tools. Some use them openly, while others prefer to hide them. Ignoring these tools seems unrealistic, but banning their use also seems unwise. Therefore, I decided to take a balanced approach, which is to openly integrate AI and observe the results. I asked students to use AI tools to help them write short argumentative paragraph drafts. The instructions were simple: write a draft, revise it, and then submit the final version along with a brief reflection on how AI played a role in their process. The results were impressive on the surface. Most of the writing was much more organized than previous assignments. Errors were reduced, cohesion improved, and students showed a higher level of confidence. If we measure success solely by the quality of writing, it can be said that this experiment was a success. This is in line with research showing that AI-supported feedback can improve students' writing, including grammar and organization, when compared to traditional methods (Sari & Lestari 2025). However, teaching English is not just about the end result. It is also related to understanding, independence, and learning progress.

The Hidden Cost of Fluency

When reading students' reflections, I began to notice a pattern. Many wrote, “AI corrected my grammar,” or “AI gave me better word choices.” Only a handful could explain the reasoning behind specific corrections. Even fewer could produce similar sentences without the tool's help.

Fluency exists, but dependence is also evident.

Connecting these results to broader research, a recent study in Indonesia shows that the use of AI-based writing tools can increase dependence and produce uniform writing in EFL learners, which ultimately has the potential to weaken independence in language learning and critical thinking skills (Zahra & Fithriani, 2025). This finding really reflects what I see in class. Students are good at using tools, but they don't necessarily internalize the language structures they produce. A critic might think this concern is overblown. Can't we allow the use of dictionaries, spell checkers, and translation tools? Isn't AI just a more advanced form of the same kind of help? The difference I see lies in how visible the thought process is. Traditional tools help in certain aspects, such as finding the meaning of a word or checking spelling. AI, on the other hand, is capable of creating complete texts that appear coherent, consistent, and accurate. This thought process becomes invisible, even to the students themselves.

Redesigning Assignments, Not Blaming Tools

Now I can be defensive by implementing strict rules, increasing supervision, or imposing severe penalties. But doing all that won't solve the root of the problem. The point is that the problem isn't with AI, but with how I design the assignment. So, I changed my approach. Instead of asking students to collect their texts or paragraphs, I asked them to document their interactions when using AI. They needed to explain the instructions they gave the reasons for accepting or rejecting suggestions from AI, rewrite paragraphs without the help of AI, and reflect on which parts they still found difficult. And suddenly, the atmosphere in the classroom changed. The students began to ask more in-depth questions, not just “Is this correct?” but “Why does the use of this word or sentence seem more academic?” or “What makes me choose this word or sentence structure?” . AI is now not just a machine that provides answers, but a dialogue partner. Unfortunately, their writing became irregular. But with that, the learning became clearer and more structured.

My Changing Role as a Teacher

Perhaps the most significant change did not occur in the students, but in myself. Before the advent of AI, I considered my role to be primarily that of a language information provider and proofreader. With AI in the classroom, that role began to feel less relevant in some ways. The tool could make corrections faster than I could. However, AI does not have the ability to understand the learning process, analyze misunderstandings, or provide more in-depth guidance. My role shifted from “corrector” to “mediator.” I no longer compete with AI, but function to help students understand and interpret it. This change was quite uncomfortable at first. It challenged my professional identity. If a machine can produce better sentences than my students or even exceed my own abilities, then what are my skills? Over time, the answer became clear. My value lies not in my ability to create English, but in my ability to teach how English works, when language choices become critical, and why context cannot be fully automated.

Lessons Learned the Hard Way

Looking back, I realize that my initial enthusiasm for AI was influenced by an unseen tendency: the belief that better results mean more effective learning. That belief is not always accurate. From this experience, I learned several important lessons. First, AI reinforces the impact of task design. Poorly designed tasks become shortcuts, while well-designed tasks create opportunities for learning. Second, openness to receiving information is more effective. Third, the influence of a teacher does not disappear; it simply shifts. Now, that control lies in evaluation, organization, and guidance, not just mechanical correction. Finally, literacy in the context of AI has now merged with language literacy. Teaching English today means equipping students to critically interact with tools that can generate English for them. Recent research by Hwang et al. (2023) supports this new idea of literacy. Several studies have introduced the term “prompt literacy,” which is the ability to create effective prompts for AI tools, as a crucial skill for learners in the AI era. This suggests that future EFL teaching approaches need to help students understand how AI “thinks,” not just what it can produce.

Moving Forward with Caution and Clarity